Using SPSS to Understand Research and Data Analysis. | |||
|
Chapter 7 7.1 Introduction to Crosstabs So far we have discussed distributions of scores on a single variable. We now turn our attention to the topic of bivariate distributions - ones which characterize relationships between two variables simultaneously.
These types of analyses examine variability in two or more distributions of scores to determine whether or not there is any pattern of covariation, or commonality, between the variables. In Chapter 8 we will discuss bivariate relationships between quantitative (continuous) variables. In the present chapter we consider relationships between qualitative (categorical) variables. Much research involves data that are categorical rather than continuous. In fact, the variable transformations we described in Chapter 5 involved converting continuous variables (e.g., masctot, on which employees could score anywhere between 5 and 35) into dichotomous categorical variables (e.g., masc, in which employees were placed into just two groups - high vs. low masculinity). We can examine relationships between categorical variables via an extension of frequency analysis. We could use the Frequencies procedure to construct a frequency table of the number of EZ employees who fall into the low vs. the high masculinity categories. However, a more interesting (and potentially more important) question we could ask is whether the frequency distribution of low vs. high masculinity is similar or different for male vs. female employees. We could generate separate frequency tables of this variable for each sex to address this question, but that would be unsystematic. Further, SPSS includes a procedure that is specifically designed to generate frequency distributions for two variables simulatneously. This procedure is called Crosstabs, and it produces tables referred to as crosstabulations, because they tabulate or count the frequencies of values across two variables simultaneously. Thus, Crosstabs allows us to answer questions such as whether there is a relationship between masculinity level and gender. A reasonable hypothesis, for example, might be that male employees are more likely than female employees to be in the high masculinity category, and female employees are more likely than male employees to be in the low masculinity category. We could test this hypothesis by examining the frequencies obtained in a crosstabulation of the variables masc and gender. Our printout would contain a table that indicates:
This will allow us to compare the number of men and women in the low/high masculinity categories, but it will also allow us to examine frequencies within gender. For example, the table would also permit us to examine whether women are more likely to be in the low-masculinity catagory than in the high-masculinity category, and vice-versa for men. Further, Crosstabs can calculate a statistic that allow us to determine whether any observed relationship between these variables is a statistically significant one or due to chance. There are numerous other questions relevant to our project that could be addressed using Crosstabs to generate bivariate frequency distributions (for example, whether the high-masculinity category is most frequently associated with the high-task skill category of leader style). You're encouraged to explore these relationships (recall that we have several categorical variables in the ezdata.sav file: gender, masc, fem, task, soc). We will use the masc and gender example to illustrate the Crosstabs procedure in this chapter, and you will be asked to do the same analysis for fem and gender in the exercise at the end of the chapter. |
|||